China sets rules for phones with pre-installed AI agents

China's Ministry of Industry and Information Technology (MIIT) published Notice No. 42 (2026) on November 6, 2026, bringing phones, tablets and wearables with pre-installed AI agents under the country's telecom network-access rules.

The ministry defines an "agent terminal" as a telecommunications device that ships with an AI agent pre-installed. Unlike an ordinary assistant, an agent interprets an instruction, accesses information, picks tools and carries out a sequence of actions — which can involve payments, messages and other user accounts.

What the notice requires

Developers must verify that the agent is compatible with specific device models and operating systems, and the product settings page must disclose the list of devices that have passed adaptation checks. Adaptation information must also be submitted to app distribution platforms before an agent is listed.

Network-access testing will now include agent compatibility and a fraud-risk assessment. The notice also covers permission use, communications functions, network and data security, user rights and how the agent performs tasks on the device.

An identity code for agents

According to analysis by the specialist newsletter Ctrl+AI+Reg, agents built into these devices will have to carry an identity code — a kind of passport for bots, making it possible to identify which agent is acting on a user's behalf. That reading comes from the newsletter, not as a literal quote from the notice.

Why it matters now

The notice arrives as AI assistants are actually reaching phones sold in China: Apple began testing Apple Intelligence in the country on November 4, with a public launch expected in December alongside iOS 27.3, using Alibaba models. The notice signals that Beijing intends to regulate AI agents as part of the device and telecom environment, not merely as software downloaded later. The topic adds to China's accelerating AI law, which requires DeepSeek, Alibaba and Moonshot to report risks and incidents.

The most important details for the market will emerge in implementation: what evidence regulators expect, how agent updates are handled, and what happens when a model or operating system changes after certification.

Image credit: photo by Dennis Cortés, via Unsplash and Wikimedia Commons (CC0).

Sources

Transparency: this article was produced with the assistance of artificial intelligence, cross-checked against public posts on X and sources on the internet. Please consult the original sources to confirm the information.

By Muse